]> git.sesse.net Git - vlc/blob - modules/packetizer/h264.c
* h264: handle SPS/PPS in annexb format in extra data (needed for RTP).
[vlc] / modules / packetizer / h264.c
1 /*****************************************************************************
2  * h264.c: h264/avc video packetizer
3  *****************************************************************************
4  * Copyright (C) 2001, 2002 the VideoLAN team
5  * $Id$
6  *
7  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
8  *          Eric Petit <titer@videolan.org>
9  *          Gildas Bazin <gbazin@videolan.org>
10  *
11  * This program is free software; you can redistribute it and/or modify
12  * it under the terms of the GNU General Public License as published by
13  * the Free Software Foundation; either version 2 of the License, or
14  * (at your option) any later version.
15  *
16  * This program is distributed in the hope that it will be useful,
17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19  * GNU General Public License for more details.
20  *
21  * You should have received a copy of the GNU General Public License
22  * along with this program; if not, write to the Free Software
23  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
24  *****************************************************************************/
25
26 /*****************************************************************************
27  * Preamble
28  *****************************************************************************/
29 #include <stdlib.h>                                      /* malloc(), free() */
30
31 #include <vlc/vlc.h>
32 #include <vlc/decoder.h>
33 #include <vlc/sout.h>
34
35 #include "vlc_block_helper.h"
36 #include "vlc_bits.h"
37
38 /*****************************************************************************
39  * Module descriptor
40  *****************************************************************************/
41 static int  Open ( vlc_object_t * );
42 static void Close( vlc_object_t * );
43
44 vlc_module_begin();
45     set_category( CAT_SOUT );
46     set_subcategory( SUBCAT_SOUT_PACKETIZER );
47     set_description( _("H264 video packetizer") );
48     set_capability( "packetizer", 50 );
49     set_callbacks( Open, Close );
50 vlc_module_end();
51
52
53 /****************************************************************************
54  * Local prototypes
55  ****************************************************************************/
56 static block_t *Packetize( decoder_t *, block_t ** );
57 static block_t *PacketizeAVC1( decoder_t *, block_t ** );
58
59 struct decoder_sys_t
60 {
61     block_bytestream_t bytestream;
62
63     int     i_state;
64     int     i_offset;
65     uint8_t startcode[4];
66
67     vlc_bool_t b_slice;
68     block_t    *p_frame;
69
70     vlc_bool_t   b_sps;
71     vlc_bool_t   b_pps;
72
73     /* avcC data */
74     int i_avcC_length_size;
75
76     /* Useful values of the Sequence Parameter Set */
77     int i_log2_max_frame_num;
78     int b_frame_mbs_only;
79
80     /* Useful values of the Slice Header */
81     int i_nal_type;
82     int i_nal_ref_idc;
83     int i_idr_pic_id;
84     int i_frame_num;
85 };
86
87 enum
88 {
89     STATE_NOSYNC,
90     STATE_NEXT_SYNC,
91 };
92
93 enum nal_unit_type_e
94 {
95     NAL_UNKNOWN = 0,
96     NAL_SLICE   = 1,
97     NAL_SLICE_DPA   = 2,
98     NAL_SLICE_DPB   = 3,
99     NAL_SLICE_DPC   = 4,
100     NAL_SLICE_IDR   = 5,    /* ref_idc != 0 */
101     NAL_SEI         = 6,    /* ref_idc == 0 */
102     NAL_SPS         = 7,
103     NAL_PPS         = 8,
104     NAL_AU_DELIMITER= 9
105     /* ref_idc == 0 for 6,9,10,11,12 */
106 };
107
108 enum nal_priority_e
109 {
110     NAL_PRIORITY_DISPOSABLE = 0,
111     NAL_PRIORITY_LOW        = 1,
112     NAL_PRIORITY_HIGH       = 2,
113     NAL_PRIORITY_HIGHEST    = 3,
114 };
115
116 static block_t *ParseNALBlock( decoder_t *, block_t * );
117
118 static block_t *nal_get_annexeb( decoder_t *, uint8_t *p, int );
119
120 /*****************************************************************************
121  * Open: probe the packetizer and return score
122  *****************************************************************************/
123 static int Open( vlc_object_t *p_this )
124 {
125     decoder_t     *p_dec = (decoder_t*)p_this;
126     decoder_sys_t *p_sys;
127
128     if( p_dec->fmt_in.i_codec != VLC_FOURCC( 'h', '2', '6', '4') &&
129         p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') &&
130         p_dec->fmt_in.i_codec != VLC_FOURCC( 'V', 'S', 'S', 'H') &&
131         p_dec->fmt_in.i_codec != VLC_FOURCC( 'v', 's', 's', 'h') &&
132         ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') ||
133           p_dec->fmt_in.i_extra < 7 ) )
134     {
135         return VLC_EGENERIC;
136     }
137
138     /* Allocate the memory needed to store the decoder's structure */
139     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
140     {
141         msg_Err( p_dec, "out of memory" );
142         return VLC_EGENERIC;
143     }
144     p_sys->i_state = STATE_NOSYNC;
145     p_sys->i_offset = 0;
146     p_sys->startcode[0] = 0;
147     p_sys->startcode[1] = 0;
148     p_sys->startcode[2] = 0;
149     p_sys->startcode[3] = 1;
150     p_sys->bytestream = block_BytestreamInit( p_dec );
151     p_sys->b_slice = VLC_FALSE;
152     p_sys->p_frame = NULL;
153     p_sys->b_sps   = VLC_FALSE;
154     p_sys->b_pps   = VLC_FALSE;
155
156     p_sys->i_nal_type = -1;
157     p_sys->i_nal_ref_idc = -1;
158     p_sys->i_idr_pic_id = -1;
159     p_sys->i_frame_num = -1;
160
161     /* Setup properties */
162     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
163     p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
164     /* FIXME: FFMPEG isn't happy at all if you leave this */
165     if( p_dec->fmt_out.i_extra ) free( p_dec->fmt_out.p_extra );
166     p_dec->fmt_out.i_extra = 0; p_dec->fmt_out.p_extra = 0;
167
168     if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
169     {
170         uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
171         int i_sps, i_pps;
172         int i;
173
174         /* Parse avcC */
175         p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
176
177         /* Read SPS */
178         i_sps = (*p++)&0x1f;
179
180         for( i = 0; i < i_sps; i++ )
181         {
182             int i_length = GetWBE( p );
183             block_t *p_sps = nal_get_annexeb( p_dec, p + 2, i_length );
184
185             ParseNALBlock( p_dec, p_sps );
186             p += 2 + i_length;
187         }
188         /* Read PPS */
189         i_pps = *p++;
190         for( i = 0; i < i_pps; i++ )
191         {
192             int i_length = GetWBE( p );
193             block_t *p_pps = nal_get_annexeb( p_dec, p + 2, i_length );
194
195             ParseNALBlock( p_dec, p_pps );
196             p += 2 + i_length;
197         }
198         msg_Dbg( p_dec, "avcC length size=%d sps=%d pps=%d",
199                  p_sys->i_avcC_length_size, i_sps, i_pps );
200
201         /* Set callback */
202         p_dec->pf_packetize = PacketizeAVC1;
203     }
204     else
205     {
206         /* Set callback */
207         p_dec->pf_packetize = Packetize;
208
209         /* */
210         if( p_dec->fmt_in.i_extra > 0 )
211         {
212             block_t *p_init = block_New( p_dec, p_dec->fmt_in.i_extra );
213             block_t *p_pic;
214
215             memcpy( p_init->p_buffer, p_dec->fmt_in.p_extra,
216                     p_dec->fmt_in.i_extra );
217
218             while( ( p_pic = Packetize( p_dec, &p_init ) ) )
219             {
220                 /* Should not occur because we should only receive SPS/PPS */
221                 block_Release( p_pic );
222             }
223         }
224     }
225
226     return VLC_SUCCESS;
227 }
228
229 /*****************************************************************************
230  * Close: clean up the packetizer
231  *****************************************************************************/
232 static void Close( vlc_object_t *p_this )
233 {
234     decoder_t *p_dec = (decoder_t*)p_this;
235     decoder_sys_t *p_sys = p_dec->p_sys;
236
237     block_BytestreamRelease( &p_sys->bytestream );
238     free( p_sys );
239 }
240
241 /****************************************************************************
242  * Packetize: the whole thing
243  ****************************************************************************/
244 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
245 {
246     decoder_sys_t *p_sys = p_dec->p_sys;
247     block_t       *p_pic;
248
249     if( !pp_block || !*pp_block ) return NULL;
250
251     block_BytestreamPush( &p_sys->bytestream, *pp_block );
252
253     for( ;; )
254     {
255         switch( p_sys->i_state )
256         {
257             case STATE_NOSYNC:
258                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
259                       &p_sys->i_offset, p_sys->startcode+1, 3 ) == VLC_SUCCESS)
260                 {
261                     p_sys->i_state = STATE_NEXT_SYNC;
262                 }
263
264                 if( p_sys->i_offset )
265                 {
266                     block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
267                     p_sys->i_offset = 0;
268                     block_BytestreamFlush( &p_sys->bytestream );
269                 }
270
271                 if( p_sys->i_state != STATE_NEXT_SYNC )
272                 {
273                     /* Need more data */
274                     return NULL;
275                 }
276
277                 p_sys->i_offset = 1; /* To find next startcode */
278
279             case STATE_NEXT_SYNC:
280                 /* Find the next startcode */
281                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
282                       &p_sys->i_offset, p_sys->startcode, 3 ) != VLC_SUCCESS)
283                 {
284                     if( block_FindStartcodeFromOffset( &p_sys->bytestream,
285                           &p_sys->i_offset, p_sys->startcode+1, 3 ) !=
286                         VLC_SUCCESS )
287                     {
288                         /* Need more data */
289                         return NULL;
290                     }
291                 }
292
293                 /* Get the new fragment and set the pts/dts */
294                 p_pic = block_New( p_dec, p_sys->i_offset );
295                 p_pic->i_pts = p_sys->bytestream.p_block->i_pts;
296                 p_pic->i_dts = p_sys->bytestream.p_block->i_dts;
297
298                 block_GetBytes( &p_sys->bytestream, p_pic->p_buffer,
299                                 p_pic->i_buffer );
300
301                 p_sys->i_offset = 0;
302
303                 /* Parse the NAL */
304                 if( !( p_pic = ParseNALBlock( p_dec, p_pic ) ) )
305                 {
306                     p_sys->i_state = STATE_NOSYNC;
307                     break;
308                 }
309 #if 0
310                 msg_Dbg( p_dec, "pts="I64Fd" dts="I64Fd,
311                          p_pic->i_pts, p_pic->i_dts );
312 #endif
313
314                 /* So p_block doesn't get re-added several times */
315                 *pp_block = block_BytestreamPop( &p_sys->bytestream );
316
317                 p_sys->i_state = STATE_NOSYNC;
318
319                 return p_pic;
320         }
321     }
322 }
323
324 /****************************************************************************
325  * PacketizeAVC1: the whole thing
326  ****************************************************************************/
327 static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
328 {
329     decoder_sys_t *p_sys = p_dec->p_sys;
330     block_t       *p_block;
331     block_t       *p_ret = NULL;
332     uint8_t       *p;
333
334     if( !pp_block || !*pp_block ) return NULL;
335
336     p_block = *pp_block;
337     *pp_block = NULL;
338
339     for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
340     {
341         block_t *p_pic;
342         int i_size = 0;
343         int i;
344
345         for( i = 0; i < p_sys->i_avcC_length_size; i++ )
346         {
347             i_size = (i_size << 8) | (*p++);
348         }
349
350         if( i_size > 0 )
351         {
352             block_t *p_part = nal_get_annexeb( p_dec, p, i_size );
353
354             p_part->i_dts = p_block->i_dts;
355             p_part->i_pts = p_block->i_pts;
356
357             /* Parse the NAL */
358             if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
359             {
360                 block_ChainAppend( &p_ret, p_pic );
361             }
362         }
363         p += i_size;
364     }
365
366     return p_ret;
367 }
368
369 static block_t *nal_get_annexeb( decoder_t *p_dec, uint8_t *p, int i_size )
370 {
371     block_t *p_nal;
372
373     p_nal = block_New( p_dec, 3 + i_size );
374
375     /* Add start code */
376     p_nal->p_buffer[0] = 0x00;
377     p_nal->p_buffer[1] = 0x00;
378     p_nal->p_buffer[2] = 0x01;
379
380     /* Copy nalu */
381     memcpy( &p_nal->p_buffer[3], p, i_size );
382
383     return p_nal;
384 }
385
386 static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret,
387                              uint8_t *src, int i_src )
388 {
389     uint8_t *end = &src[i_src];
390     uint8_t *dst = malloc( i_src );
391
392     *pp_ret = dst;
393
394     while( src < end )
395     {
396         if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00 &&
397             src[2] == 0x03 )
398         {
399             *dst++ = 0x00;
400             *dst++ = 0x00;
401
402             src += 3;
403             continue;
404         }
405         *dst++ = *src++;
406     }
407
408     *pi_ret = dst - *pp_ret;
409 }
410
411 static inline int bs_read_ue( bs_t *s )
412 {
413     int i = 0;
414
415     while( bs_read1( s ) == 0 && s->p < s->p_end && i < 32 )
416     {
417         i++;
418     }
419     return( ( 1 << i) - 1 + bs_read( s, i ) );
420 }
421
422 static inline int bs_read_se( bs_t *s )
423 {
424     int val = bs_read_ue( s );
425
426     return val&0x01 ? (val+1)/2 : -(val/2);
427 }
428
429
430 static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
431 {
432     decoder_sys_t *p_sys = p_dec->p_sys;
433     block_t *p_pic = NULL;
434
435     const int i_nal_ref_idc = (p_frag->p_buffer[3] >> 5)&0x03;
436     const int i_nal_type = p_frag->p_buffer[3]&0x1f;
437
438 #define OUTPUT \
439     do {                                                \
440         p_pic = block_ChainGather( p_sys->p_frame );    \
441         p_pic->i_length = 0;    /* FIXME */             \
442                                                         \
443         p_sys->p_frame = NULL;                          \
444         p_sys->b_slice = VLC_FALSE;                     \
445     } while(0)
446
447
448     if( p_sys->b_slice && !p_sys->b_sps )
449     {
450         block_ChainRelease( p_sys->p_frame );
451         msg_Warn( p_dec, "waiting for SPS" );
452
453         /* Reset context */
454         p_sys->p_frame = NULL;
455         p_sys->b_slice = VLC_FALSE;
456     }
457
458     if( !p_sys->b_sps &&
459         i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
460     {
461         p_sys->b_slice = VLC_TRUE;
462         /* Fragment will be discarded later on */
463     }
464     else if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
465     {
466         uint8_t *dec;
467         int i_dec, i_first_mb, i_slice_type, i_frame_num, i_pic_flags = 0;
468         vlc_bool_t b_pic = VLC_FALSE;
469         bs_t s;
470
471         /* do not convert the whole frame */
472         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
473                          __MIN( p_frag->i_buffer - 4, 60 ) );
474         bs_init( &s, dec, i_dec );
475
476         /* first_mb_in_slice */
477         i_first_mb = bs_read_ue( &s );
478
479         /* slice_type */
480         switch( (i_slice_type = bs_read_ue( &s )) )
481         {
482             case 0: case 5:
483                 i_pic_flags = BLOCK_FLAG_TYPE_P;
484                 break;
485             case 1: case 6:
486                 i_pic_flags = BLOCK_FLAG_TYPE_B;
487                 break;
488             case 2: case 7:
489                 i_pic_flags = BLOCK_FLAG_TYPE_I;
490                 break;
491             case 3: case 8: /* SP */
492                 i_pic_flags = BLOCK_FLAG_TYPE_P;
493                 break;
494             case 4: case 9:
495                 i_pic_flags = BLOCK_FLAG_TYPE_I;
496                 break;
497         }
498
499         /* pic_parameter_set_id */
500         bs_read_ue( &s );
501         /* frame_num */
502         i_frame_num = bs_read( &s, p_sys->i_log2_max_frame_num + 4 );
503
504         /* Detection of the first VCL NAL unit of a primary coded picture
505          * (cf. 7.4.1.2.4) */
506         if( i_frame_num != p_sys->i_frame_num ||
507             ( (i_nal_ref_idc != p_sys->i_nal_ref_idc) &&
508               (!i_nal_ref_idc || !p_sys->i_nal_ref_idc) ) )
509         {
510             b_pic = VLC_TRUE;
511         }
512         p_sys->i_frame_num = i_frame_num;
513         p_sys->i_nal_ref_idc = i_nal_ref_idc;
514
515         if( !p_sys->b_frame_mbs_only )
516         {
517             /* field_pic_flag */
518             if( bs_read( &s, 1 ) )
519             {
520                 /* bottom_field_flag */
521                 bs_read( &s, 1 );
522             }
523         }
524
525         if( i_nal_type == NAL_SLICE_IDR )
526         {
527             /* id_pic_id */
528             int i_idr_pic_id = bs_read_ue( &s );
529             if( p_sys->i_nal_type != i_nal_type ) b_pic = VLC_TRUE;
530             if( p_sys->i_idr_pic_id != i_idr_pic_id ) b_pic = VLC_TRUE;
531             p_sys->i_idr_pic_id = i_idr_pic_id;
532         }
533         p_sys->i_nal_type = i_nal_type;
534
535         if( b_pic && p_sys->b_slice )
536             OUTPUT;
537
538         p_sys->b_slice = VLC_TRUE;
539
540         free( dec );
541     }
542     else if( i_nal_type == NAL_SPS )
543     {
544         uint8_t *dec;
545         int     i_dec;
546         bs_t s;
547         int i_tmp;
548
549         if( !p_sys->b_sps )
550             msg_Dbg( p_dec, "found NAL_SPS" );
551
552         p_sys->b_sps = VLC_TRUE;
553
554         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
555                          p_frag->i_buffer - 4 );
556
557         bs_init( &s, dec, i_dec );
558         /* Skip profile(8), constraint_set012, reserver(5), level(8) */
559         bs_skip( &s, 8 + 1+1+1 + 5 + 8 );
560         /* sps id */
561         bs_read_ue( &s );
562         /* Skip i_log2_max_frame_num */
563         p_sys->i_log2_max_frame_num = bs_read_ue( &s );
564         /* Read poc_type */
565         i_tmp = bs_read_ue( &s );
566         if( i_tmp == 0 )
567         {
568             /* skip i_log2_max_poc_lsb */
569             bs_read_ue( &s );
570         }
571         else if( i_tmp == 1 )
572         {
573             int i_cycle;
574             /* skip b_delta_pic_order_always_zero */
575             bs_skip( &s, 1 );
576             /* skip i_offset_for_non_ref_pic */
577             bs_read_se( &s );
578             /* skip i_offset_for_top_to_bottom_field */
579             bs_read_se( &s );
580             /* read i_num_ref_frames_in_poc_cycle */
581             i_cycle = bs_read_ue( &s );
582             if( i_cycle > 256 ) i_cycle = 256;
583             while( i_cycle > 0 )
584             {
585                 /* skip i_offset_for_ref_frame */
586                 bs_read_se(&s );
587             }
588         }
589         /* i_num_ref_frames */
590         bs_read_ue( &s );
591         /* b_gaps_in_frame_num_value_allowed */
592         bs_skip( &s, 1 );
593
594         /* Read size */
595         p_dec->fmt_out.video.i_width  = 16 * ( bs_read_ue( &s ) + 1 );
596         p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
597
598         /* b_frame_mbs_only */
599         p_sys->b_frame_mbs_only = bs_read( &s, 1 );
600         if( p_sys->b_frame_mbs_only == 0 )
601         {
602             bs_skip( &s, 1 );
603         }
604         /* b_direct8x8_inference */
605         bs_skip( &s, 1 );
606
607         /* crop */
608         i_tmp = bs_read( &s, 1 );
609         if( i_tmp )
610         {
611             /* left */
612             bs_read_ue( &s );
613             /* right */
614             bs_read_ue( &s );
615             /* top */
616             bs_read_ue( &s );
617             /* bottom */
618             bs_read_ue( &s );
619         }
620
621         /* vui */
622         i_tmp = bs_read( &s, 1 );
623         if( i_tmp )
624         {
625             /* read the aspect ratio part if any FIXME check it */
626             i_tmp = bs_read( &s, 1 );
627             if( i_tmp )
628             {
629                 static const struct { int w, h; } sar[14] =
630                 {
631                     { 0,   0 }, { 1,   1 }, { 12, 11 }, { 10, 11 },
632                     { 16, 11 }, { 40, 33 }, { 24, 11 }, { 20, 11 },
633                     { 32, 11 }, { 80, 33 }, { 18, 11 }, { 15, 11 },
634                     { 64, 33 }, { 160,99 },
635                 };
636                 int i_sar = bs_read( &s, 8 );
637                 int w, h;
638
639                 if( i_sar < 14 )
640                 {
641                     w = sar[i_sar].w;
642                     h = sar[i_sar].h;
643                 }
644                 else
645                 {
646                     w = bs_read( &s, 16 );
647                     h = bs_read( &s, 16 );
648                 }
649                 if( h != 0 )
650                     p_dec->fmt_out.video.i_aspect =
651                         VOUT_ASPECT_FACTOR * w / h * p_dec->fmt_out.video.i_width /
652                         p_dec->fmt_out.video.i_height;
653                 else
654                     p_dec->fmt_out.video.i_aspect = VOUT_ASPECT_FACTOR;
655             }
656         }
657
658         free( dec );
659
660
661         if( p_sys->b_slice )
662             OUTPUT;
663     }
664     else if( i_nal_type == NAL_PPS )
665     {
666         bs_t s;
667         bs_init( &s, &p_frag->p_buffer[4], p_frag->i_buffer - 4 );
668
669         if( !p_sys->b_pps )
670             msg_Dbg( p_dec, "found NAL_PPS" );
671         p_sys->b_pps = VLC_TRUE;
672
673         /* TODO */
674
675         if( p_sys->b_slice )
676             OUTPUT;
677     }
678     else if( i_nal_type == NAL_AU_DELIMITER ||
679              i_nal_type == NAL_SEI ||
680              ( i_nal_type >= 13 && i_nal_type <= 18 ) )
681     {
682         if( p_sys->b_slice )
683             OUTPUT;
684     }
685
686 #undef OUTPUT
687
688     /* Append the block */
689     block_ChainAppend( &p_sys->p_frame, p_frag );
690
691     return p_pic;
692 }